a

r
X
i
v
:
0
9
1
0
.
1
5
8
9
v
3


[
h
e
p
-
t
h
]


1

A
p
r

2
0
1
0
How many universes are in the multiverse?
Andrei Linde

and Vitaly Vanchurin

Department of Physics, Stanford University, Stanford, CA 94305, USA

(Dated: April 2, 2010)
We argue that the total number of distinguishable locally Friedmann universes generated by
eternal ination is proportional to the exponent of the entropy of inationary perturbations and is
limited by e
e
3N
, where N is the number of e-folds of slow-roll post-eternal ination. For simplest
models of chaotic ination, N is approximately equal to de Sitter entropy at the end of eternal
ination; it can be exponentially large. However, not all of these universes can be observed by a
local observer. In the presence of a cosmological constant the number of distinguishable universes
is bounded by e
||
3/4
. In the context of the string theory landscape, the overall number of dierent
universes is expected to be exponentially greater than the total number of vacua in the landscape.
We discuss the possibility that the strongest constraint on the number of distinguishable universes
may be related not to the properties of the multiverse but to the properties of observers.
I. INTRODUCTION
With the invention of inationary cosmology, the no-
tion of a uniform universe was gradually replaced by the
notion of a multiverse consisting of many locally uniform
exponentially large parts [1, 2]. Each of these parts lo-
cally looks like a uniform nearly-Friedmann universe. A
collection of all of these universes represents an eternally
growing fractal consisting of many such universes with
dierent properties [35]. This scenario recently became
quite popular when a mechanism to stabilize string the-
ory vacua was found [6], and string theorists realized [7],
in agreement with earlier expectations [8, 9], that the to-
tal number of dierent stringy vacua can be extremely
large. The popular estimate of the number of dierent
vacua is 10
500
, but the true number may be much
smaller or much greater than that [7]. Because of the
transitions from one vacuum state to another, the ina-
tionary multiverse becomes divided into an exponentially
large number of dierent exponentially large universes
with dierent laws of low-energy physics operating in
each of them. This picture, which is now known as the
string theory landscape [10], was envisaged in the very
rst paper on eternal chaotic ination [4].
But the string theory landscape does not fully describe
all of the options which may exist in an inationary mul-
tiverse. The properties of our world are determined not
only by the properties of the vacuum state; after all, we
do not live in a vacuum. We live in the world with a
certain matter content and a particular large-scale struc-
ture. Even if we concentrate on a single vacuum state,
i.e. on the same position in the string theory landscape,
the large-scale structure (e.g. the spatial distribution of
galaxies) and the matter content in each of the locally

Electronic address: [email protected]

Electronic address: [email protected]

Friedmann parts of the cosmic fractal may be quite dier-
ent. One may wonder how many dierent locally Fried-
mann universes one may encounter in any particular part
of the landscape. Yet another question is how many lo-
cally distinguishable classical geometries of the universe
one may encounter inside our cosmological horizon.
Certain aspects of the multiplicity of the universe in
the context of eternal ination were previously discussed
in Ref. [11]. The authors were mainly interested in
the number of possible outcomes which may appear in-
side a horizon-size classically uniform part of a post-
recombination universe because of quantum or thermal
uctuations. Our goals are somewhat dierent: we will
evaluate the number of inhomogeneous but locally Fried-
mann parts of the multiverse, which are dierent at a
classical level. In a certain sense, which is going to be
clear from the subsequent discussion, we will calculate
the total number of classical histories for the geometry of
the universe. More precisely, we calculate the total num-
ber of dierent coarse grained classical post-inationary
initial conditions, which determine the subsequent evo-
lution of the universe.
Our work is a part of the general eort towards nding
the probability to live in a universe with some particular
set of properties. This requires nding the probability
measure in the multiverse, see e.g. [1224]. However,
before embarking on that grand quest, it may be useful to
solve a simpler problem: to classify and count all possible
universes. This is the main goal of our paper.
The paper is organized as follows. In Sec. II we esti-
mate the entropy of cosmological perturbations generated
during slow-roll ination and in Sec. III we calculate the
total number of distinguishable universes produced by
eternal ination. The number of universes in the pres-
ence of cosmological constant is evaluated in Secs. IV
and in the context of the string theory landscape in Sec.
V. In Sec. VI we will briey discuss entropy of normal
matter, even though this entropy is not directly related
to the number of possible classical geometries of the uni-
2
verse. In Sec. VII we argue that only a small fraction of
all universes can even in principle be distinguished by a
local observer. The main results are summarized in the
Sec. VIII.
II. COUNTING THE UNIVERSES
When one starts thinking about the total number of
various possibilities which exist in the multiverse, the rst
idea that comes to mind is to take into account all quan-
tum uctuations, or real particles in a state of thermody-
namical equilibrium, which are present even in a totally
homogeneous universe [11]. For example, one could argue
that the total number of all possible quantum congura-
tions inside an event horizon of size H
1

1/2
in de
Sitter space with the cosmological constant is given by
e
S
dS
, where S
dS
= 24
2

1
is the Gibbons-Hawking dS
entropy. Those who would like to deal with something
more tangible than quantum uctuations, could suggest
to calculate the quantity e
Smatter
, describing the entropy
of all particles in a given part of the universe. How-
ever, in this paper we are not going to count short living
uctuations, or Boltzmann brains, which may emerge be-
cause of exponentially improbable quantum (or thermal)
uctuations even in an absolutely uniform space without
planets and galaxies. Thus we should turn our attention
to something else.
According to inationary cosmology, the large-scale
structure of the universe, which is necessary for our ex-
istence, is a result of quantum eects which occurred at
the stage of slow-roll ination [25]. This is a distinguish-
ing feature of the slow-roll ination, as compared to the
false vacuum ination in dierent metastable dS vacua
of the landscape.
During ination with the Hubble constant H
I
, quan-
tum uctuations of all scalar elds with masses m < H
I
are generated. These perturbations produced during a
typical time H
1
I
have a typical amplitude =
HI
2
and a wavelength O(H
1
). When they are stretched to
an exponentially large scale by ination, they stop oscil-
lating (freeze) and start looking as a nearly homogeneous
classical scalar eld. With each new e-fold of ination,
new perturbations are generated on top of the previously
generated ones. This is the standard mechanism of pro-
duction of perturbations of metric responsible for forma-
tion of the large scale structure of the universe. In this
manner, quantum uctuations during ination prepare
dierent classical initial conditions for the subsequent
evolution of dierent parts of the universe.
Note that this process occurs independently in each
part of the universe of size H
1
I
. Classical scalar elds
produced by freezing and stretching of inationary quan-
tum uctuations determine classical initial conditions for
all physical processes in the post-inationary universe, on
an exponentially large scale. The properties of quantum
jumps determine the properties of the universe on a scale
corresponding to the size of the initial Hubble size domain
stretched by the subsequent cosmological evolution.
If the inaton eld driving ination jumps in the di-
rection opposite to its classical motion (i.e. uphill), this
produces a slightly overdense region on the corresponding
scale; if the eld jumps downhill, it produces an under-
dense region. In other words, geometric properties of our
world are determined by the chain reaction of quantum
jumps during ination.
Once ination is over, it leaves its remnants in form
of the large scale classical perturbations of metric and
of various light scalar elds. If one plugs in all of these
perturbations in a powerful computer and evolve them in
accordance with the laws of physics operating in our part
of the landscape, one should be able, at least in principle,
to nd all details of the subsequent galaxy formation, star
formation, post-inationary entropy production, etc.
In other words, one can argue that the evolution and
properties of all important macroscopic features of the
universe can be traced back to the two main ingredients:
1) The properties of our vacuum state, represented by
one of the many vacua in the landscape.
2) The properties of the slow-roll ination and of the
large scale perturbations of metric and physical elds pro-
duced at that stage.
The main idea of our paper is that once one nds all
possible combinations of these two ingredients, one can
determine all classical histories of the universe, and, con-
sequently, all possible macroscopic features of our uni-
verse that are required for our existence.
There are some obvious exceptions from this conjec-
ture. For example, if the slow roll ination is very short,
we may be aected by bubble collisions or other physical
processes preceding the stage of the slow roll ination. In
what follows, we will assume that the slow roll ination
is suciently long to allow us to ignore these processes,
but in general on should take them into account. Also,
the slow roll ination is not the only process which gen-
erates classical elds in cosmology; another important
example is preheating after ination [26]. The dierence
between these two processes is that the growth of the
occupation numbers of quantum uctuations during the
post-inationary preheating has a power-law dependence
on masses and coupling constants, whereas the growth of
occupation numbers of particles during ination is expo-
nentially large. That is why we will concentrate on this
process in our paper.
Now we will make an estimate of the total number of
dierent classical geometries which may be produced by
the slow-roll ination. In our estimates we will make
an important simplifying assumption. We will assume
that the eld make a single jump each time H
1
I
, and
the magnitude of the jump is
HI
2
. In other words, we
consider coarse-grained histories, ignoring, e.g., the pos-
3
sibility that the eld may, with an exponentially small
probability, jump up or down by much more than
HI
2
.
1
It is rather straightforward to calculate the total num-
ber of dierent coarse-grained geometries produced by
this mechanism. Consider an inationary domain of ini-
tial size H
1
I
after it experienced N e-folds of ination.
After the rst e-fold the domain has grown e times, and
it contains now e
3
domains of size H
1
I
in each of which
the eld could independently jump either by +
HI
2
or by

HI
2
. The total number of dierent coarse-grained con-
gurations of the eld in this domain becomes 2
e
3
. Note
that our estimate was very rough. We assumed that the
eld could experience only two possible jumps
HI
2
and
nothing in between (coarse-graining). Therefore our es-
timate is valid only up to a factor O(1) in the exponent,
so one can write the nal result as e
C e
3
, with C = O(1).
During the next time interval H
1
I
each of the e
3
do-
mains of size H
1
I
experience a similar set of jumps. They
change the value of the scalar eld on the scale H
1
I
, but
do not change the results of the previous jumps on the
scale eH
1
I
. The total number of dierent eld congu-
rations becomes e
C(e
3
+e
6
)
. Obviously, the total number
of dierent congurations after N e-folds of ination be-
comes
N exp
_
C
N

1
e
3N
_
= exp
_
ce
3N
_
, (1)
where c is another constant O(1). In what follows, we
will write the nal result in a simplied way,
N e
e
3N
, (2)
keeping in mind the uncertainty in the coecient in the
exponent. Of course, for very large N not all of these
dierent universes will be seen by any particular observer;
see a discussion of this issue in Section IV. The estimate
given above describes the total number of all possible
universes which can be seen by all possible observers.
Note that the derivation of this result has a transpar-
ent physical interpretation: Each of the e
3N
independent
inationary domains has its own degree of freedom (the
inaton eld inside it jumps either up or down). This
suggests that, up to a numerical factor, this system of
long-wavelength perturbations produced during the slow-
roll ination has entropy S
in
e
3N
. As we will see this
entropy could be associated with the entropy of cosmo-
logical perturbations derived in Refs. [3742]. We should
1
This is a delicate issue since such trajectories may be important
with some of the probability measures [13]. However, such mea-
sures suer from the youngness paradox, so we will ignore this
issue in our paper.
note that this entropy is totally dierent from the stan-
dard de Sitter entropy and from the entropy of normal
matter.
To get a numerical estimate, suppose that our part of
the universe was produced as a result of 60 e-folds of the
slow-roll ination of an inationary domain of size H
1
I
.
2
This process may create
N e
e
180
10
10
77
(3)
universes with dierent geometrical properties. This
number is incomparably greater than 10
500
. If the ini-
tial size of the universe is greater than H
1
I
, the total
number of dierent universes is even much greater.
The diversity of various outcomes of inationary evo-
lution becomes even greater if there are more than one
scalar eld with the mass smaller than H
I
. In such the-
ories not only the local geometry but even the matter
content of the universe in any given vacuum may also
depend on inationary quantum uctuations. For exam-
ple, the baryon/photon ratio n
B
/n

in the Aeck-Dine
baryogenesis scenario [27] depends on perturbations of
the scalar eld responsible for CP violation, and there-
fore it may take dierent values in dierent parts of an
inationary universe [28]. The ratio of dark matter to
baryons
DM
/
B
in axion cosmology is determined by
long wavelength inationary perturbations of the axion
eld, which takes dierent values in dierent parts of the
multiverse [29, 30]. In the curvaton theory [3235], the
amplitude of perturbations of metric is dierent in dif-
ferent parts of the multiverse [32, 36]. We will return to
these possibilities later on.
In the remainder of this section we give an alternative
interpretation of our results by following the analysis of
Ref. [37]. The authors showed that both types of cos-
mological perturbations (gravitational waves and density
perturbations) can be described by a stochastic scalar
eld , whose entropy in the limit of large occupation
numbers is given by
S V
_
d
3
k log
_
n

k
_
, (4)
where n

k
is the number of particles. The spectrum of
gravitational waves
h
and density perturbations

is
usually dened through corresponding two point corre-
lation functions which could also be expressed through
the average number of particles n

k
. Therefore, it is a
straightforward exercise to estimate the number of parti-
cles from a given spectrum of cosmological perturbations
(see Refs. [38] for details).
From (4) we can approximate the entropy of gravita-
tional radiation contained inside volume V = H
3
I
e
3N
of
2
In Sec. IV we will show explicitly that the number of observ-
able e-folds must be bounded by a logarithm of the cosmological
constant (e.g. N 70 in our universe).
4
the reheating surface
S
gw
H
3
I
e
3N
_
HI
HI e
N
k
2
dk log
_

h
a
k
_
e
3N
, (5)
where H
I
is the Hubble scale during ination and N is
the number of e-folds of slow-roll ination. The integral
in (5) is dominated by the high frequency modes k H
I
indicating that most of the entropy is generated when a
given mode crosses horizon,
S
V
H
3
I
. (6)
Similarly, the entropy density of adiabatic perturbations
is given by (6) up to a logarithmic correction [38]. In the
limit of large occupation numbers the overall entropy in
linear perturbations can be expressed as
S
pert
= c e
3N
, (7)
where c is some constant of order unity. It is now conve-
nient to dene the total number of universes as
N e
Spert
= e
c e
3N
, (8)
which agrees qualitatively with our previous estimate (2).
We should note that an accurate denition of the en-
tropy of perturbations of metric requires a more detailed
discussion; we refer the readers to the original literature
on this subject, see e.g. [3742] and references therein.
For the purposes of our paper, we will use the concept of
entropy of perturbations of metric as a shortcut interpre-
tation of our original estimate (2): S
pert
log N e
3N
.
Another important comment here is that the main con-
tribution to this number is given by the perturbations
produced at the very end of ination. These perturba-
tions are only marginally classical. Therefore in order
to use our estimates in a reliable way one should make a
step back from the very end of ination. This will some-
what reduce the extremely large numbers that we are
going to discuss shortly. However, one may expect our
estimates to be qualitatively correct in the large N limit.
III. NUMBER OF UNIVERSES PRODUCED BY
ETERNAL INFLATION
If quantum jumps of the eld dominate its classi-
cal rolling during a typical time H
1
I
, then each do-
main of a size H
1
I
will eternally split into many new
domains, in some of which the eld will over and again
jump against the classical rolling of the scalar eld, for-
ever re-starting the slow-roll process in dierent H
1
I
-
sized domains. This leads to eternal ination [3, 4]. It
occurs for elds satisfying the following generic condition
[4]:

quant

class
(t = H
1
I
) =
V

3H
2
I
()
V
3
12
2
(V

)
2
. (9)
At the rst glance, one could expect that the total
number of dierent locally Friedmann universes produced
by eternal ination must be innite since in this regime
the number of e-foldings N in (2) becomes indenitely
large. However, this is not the case. Indeed, quantum
uctuations which occur in the regime of eternal ination
produce perturbations of metric which are greater than
O(1) at the end of ination [43]. One can see it directly
by comparing the condition required for eternal ination
(9) with the amplitude of post-inationary perturbations
of metric, which are of the order
V
3/2
V

. Thus, all pertur-
bations above the boundary of eternal ination produce
the universes which do not look like locally Friedmann
universes, even approximately. That is why in order to
nd all nearly Friedmann universes produced by ina-
tion it is sucient to study the cosmological evolution of
those parts of the universe where the condition (9) is not
satised and eternal ination is over.
We will denote the boundary value of the eld at which
the condition of slow roll eternal ination is satised as

. To calculate the total number of e-foldings after the

end of eternal ination in any particular part of the uni-
verse, we should take

as the initial condition for

the phase of slow-roll ination, which leads to a nite
amount of slow-roll ination [44].
In slow-roll ination the Hubble constant is given by
_
V/3, in Planck units, and
3H
I

= V

(10)
Using expression for de Sitter entropy
S = 24
2
V
1
= 8
2
H
2
I
(11)
and the relation

N = H
I
for the number of e-foldings N,
one can easily nd that
dS
dN
=
8
2

2
H
4
I

_
2
(12)
By integrating this equation, taking into account that

< 1 and assuming that dS entropy at the end of the

slow rolling is larger than at the beginning, one can get
a bound on the total number of e-foldings,
N
tot
S
end
, (13)
where S
end
is the Gibbons-Hawking de Sitter entropy at
the end of slow roll ination [48].
This is an interesting theoretical bound, but it is not
particularly informative in practical applications. Con-
sider, for example, a simple model of the type of new
ination, or ination near an inection point, with po-
tential
V = V
0
_
1

p
p

p
_
. (14)
Note that here we absorbed V
0
in the denition of
p
. To
distinguish this case from the simplest versions of chaotic
5
ination scenario involving large elds > 1, we will
assume that
p
1. In this regime, ination begins at
0 and ends at 1. In this situation the number
of e-folds after eternal ination is given by [44]
N
tot

(12
2
)
p2
2p2
p 2

1
p1
p
V

p2
2p2
0
. (15)
Consider for example the theory of the type of new ina-
tion, with V = V
0
_
1
4
4

4
_
. In this case one has
N
tot
(
4
V
0
)

1
3
. (16)
One can show that for
4
> 1 and V 1 the bound
N
tot
S
end
is satised in this scenario, but S
end
V
1
is very much dierent from the actual number of e-
foldings after the end of eternal ination.
The situation is especially interesting and instructive
in simplest models of chaotic ination with
V =

n
n

n
. (17)
In this case, the total number of e-folds since the end of
eternal ination can be estimated by
N
tot
2
2

/n 2
_
12
2

n
_
2
n+2
n
1n
n+2
. (18)
In fact, one can easily check that in this class of theories
N
tot
C
n
S
e
, (19)
where S
e
is the de Sitter entropy at the boundary of
eternal ination and
C
n
=
n
n1
n+2
4
= O(1) (20)
for the simplest chaotic ination models with n = O(1).
To give a particular numerical estimate, in the theory
m
2

2
/2,
N
tot
c m
1
. (21)
where c = 2
5/3
3
1/2
= O(20). In realistic models one
may expect m 3 10
6
, and therefore
N
tot
10
7
. (22)
Meanwhile the bound (13) in this case would be N
tot

m
2
10
11
, which is much weaker and less informa-
tive than the actual result N
tot
m
1
which we just
obtained.
The total number of dierent types of universes pro-
duced in chaotic ination with V = m
2

2
/2, m
3 10
6
, can be estimated by
N e
e
3N
tot
e
e
3c/m
10
10
10
7
(23)
This number may change signicantly if we use a dier-
ent denition of the boundary of the eternal ination [45],
but with any denition, this number is VERY large. It is
exponentially greater than the total number of string the-
ory vacua. This number may become even much greater
if we take into account that the parameters of ination-
ary models may take dierent values in dierent vacua
in the landscape.
IV. NUMBER OF UNIVERSES IN THE
PRESENCE OF THE COSMOLOGICAL
CONSTANT
Not all of the universes produced since the end of eter-
nal ination can be distinguished by observers populating
the observable part of the universe.
During the post-inationary expansion of the universe,
each domain of initial size H
1
I
grows as H
1
I
a(t), where
a(t) is the scale factor, which is normalized to 1 at the
end of ination. At this stage the total size of the ob-
servable part of the universe grows approximately as t, so
the total number of independent domains of initial size
H
1
I
accessible to observations (i.e. the total entropy of
observable cosmological perturbations) grows as
S
pert
(t)
_
tH
I
a(t)
_
3
. (24)
This regime continues only until the moment when the
energy density of all matter becomes smaller than the
absolute value of the cosmological constant .
For > 0, starting from the time t
1/2
the uni-
verse starts expanding exponentially, and we no longer
see new parts of the universe, which leads to a cuto
in the observable information stored in the cosmological
perturbations. Meanwhile for < 0 the universe typi-
cally collapses within the time t ||
1/2
. Thus in both
cases in order to estimate the total entropy of observable
cosmological perturbations it is sucient to limit our-
selves to what one can observe within the cosmological
time t ||
1/2
.
At t ||
1/2
, the energy density of gravitational
waves, which contribute only a fraction to the overall
matter density, must be strictly smaller than the abso-
lute value of the cosmological constant,

gw
= H
4
I
a
4
< || . (25)
The above bound can be saturated only if the en-
ergy density of gravitational waves dominate the energy
density of all other types of matter at the epoch when
this energy density decreases and approaches the value
comparable to ||. By combining (24), (25) and using
H =
_
|| /3 we nd
S
pert
||
3/4
. (26)
6
One should note that our estimates are valid for the
universes which have geometry not too dierent from the
geometry of a at universe, = O(1). If one considers
open universes with 1, which are very dierent from
the universe where we live now, the total entropy inside
the observable part of the universe may be much greater
than ||
3/4
, approaching
1
for > 0 and
2
for
> 0 [46]. We will not study this regime in our paper.
It follows from (7) and (26) that the maximum number
of observable e-folds is bounded by
N
max

log(||)
4
. (27)
This number is typically much smaller than the total
number of e-folds estimated in (21). For example, the
maximum number of observable e-folds in our vacuum
with 10
120
is about 70, which is pretty close to
what is actually observed.
At rst, it could seem that the bound (26) can always
be saturated regardless of the scale of ination. However,
usually this is not the case. Suppose ination ends at
a = 1 and after an instant stage of reheating the universe
becomes dominated by matter with p
w
= w
w
. Then, at
the time when the density of matter becomes comparable
to the value of the cosmological constant one has

w
= H
2
I
a
3(1+w)
|| t
2
. (28)
From (24), one nds that the maximal value of the ob-
servable entropy is
S
pert
H
1+3w
1+w
I
||

1+3w
2+2w
. (29)
In order to analyze a particular semi-realistic example,
consider the universe dominated by relativistic matter
soon after the end of ination (w = 1/3). In this case
S
pert
H
3
2
I
||
3/4
. (30)
In this regime the bound (26) is saturated if ination ends
at the Planck density, H
2
I
= O(1).
3
However, in realistic
models of ination with H
2
I
10
9
one nds S
pert

H
3
2
I
||
3/4
||
3/4
. In particular, in the simplest
chaotic ination model with V = m
2

2
/2, m 310
6
,
and 10
120
the maximal contribution to the entropy
is given by the last stage of ination where H
I
m
3 10
6
, so one nds (assuming instant reheating and
w = 1/3):
S
pert
5 10
9
||
3/4
10
82
. (31)
3
Looking at Eq. (29), one could expect that, for example, for the
sti equation of state w = 1 one could have entropy O(
1
),
which is much greater than the bound (26). However, one can
show that in this case the energy of gravitational waves eventu-
ally begins to dominate and the bound (26) holds, as it should.
which gives the total number of dierent universes
N 10
10
82
. (32)
Of course this is a very rough estimate. In particular,
as we already mentioned, the largest contribution to this
number is given by perturbations produced at the lat-
est stages of ination. Such perturbations do not have
much time to inate and their occupation numbers are
not exponentially large, unless one makes a suciently
large step back from the end of ination, which eec-
tively decreases the number of e-foldings contributing to
our estimate. Moreover, one may argue that the infor-
mation about the last few e-foldings of ination may be
erased by subsequent cosmological evolution. This may
somewhat reduce the estimated power 82 in (32), but the
total number of possible observable universes will remain
extremely large.
Before we discuss a similar result in the context of the
string theory landscape, we should note that one may be
interested not in what could be potentially possible in
the unlimited future, but in what is possible within some
important range of time. If, for example, we are inter-
ested in the total number of options for the observable
part of the universe with age t 10
10
years, then the
results will be essentially the same as in the models with
10
120
.
V. NUMBER OF UNIVERSES IN THE
LANDSCAPE
Now let us estimate the number of distinct universes in
the entire landscape. If we assume that the total number
of vacua is M, then from (26) the total number of distinct
universes is given by a sum over all vacua
N
M

i=1
e
|i|

3
4
. (33)
Here, in order to make a rough estimate, we assumed
that the upper bound (26) can be saturated. Clearly,
the largest contribution to the number of universes (33)
comes from the vacua with the smallest absolute value of
the cosmological constant.
As we already mentioned, the popular estimate for M
is 10
500
, but in fact it can be much smaller or much
greater than that. Assuming for simplicity that the vacua
are atly distributed near = 0, one may expect that
the lowest nonvanishing value of is |
min
| 1/M.
Then from our estimates it would follow that the maxi-
mal number of observable e-folds is N
max
290 and the
corresponding number of distinct universes is
N e
|min|

3
4
e
M
3
4
10
10
375
. (34)
As we see, the total number of the observable geometries
of the universe is expected to be exponentially greater
7
than the total number M of string theory vacua in the
landscape: N e
M
3
4
.
But what if the minimal value of in the landscape
is = 0? This is a viable possibility. In fact, one of
the vacua in string theory landscape, which corresponds
to the decompactied 10D universe, does have = 0.
Does this mean that an observer in such vacua will see
an innite number of universes?
The answer is that for very small we would be able to
see the universe on the scale corresponding to the maxi-
mal number of e-folding in the slow-roll regime, or on the
scale corresponding to the boundary of self-reproduction.
In the last case, the total number of dierent observable
universes will be given by N estimated in Section III.
The situation becomes a bit more subtle in an open
universe if the slow-roll evolution is not eternal. Esti-
mates given in the previous sections rely on the assump-
tion of atness O(1), meanwhile, as we already men-
tioned, for 1, one can nd an even much greater
number of distinguishable universes, with
3/4
in (33)
replaced by
1
or
2
for positive or negative values of
respectively [46]. Moreover, in the models with van-
ishing cosmological constant the observable area of the
reheating surface is not bounded from above. An ob-
server in such universe (sometimes called the census taker
[51]) could in principle count an innite number of dis-
tinguishable universes. However, the only known vacuum
with exactly zero vacuum energy is the vacuum which ap-
pears when the universe decompacties and become 10D.
In general, supersymmetric 3D vacua may exist in string
theory landscape [52], but as of now we do not see any
reason for the tremendous ne-tuning required for this to
happen. As we already mentioned, if the vacua are atly
distributed near = 0, one may expect that the low-
est nonvanishing value of is |
min
| 1/M, which was
the origin of our estimate (34) of the number of dierent
universes with = O(1): N e
M
3
4
.
In our estimates in the last two sections we made an
assumption that local properties of our universe cannot
be aected by uctuations on the scale much greater than
the present horizon. This assumption can be violated in
the theories with more than one light scalar eld. For
example, quantum uctuations of the axion eld during
ination may produce perturbations with the wavelength
many orders of magnitude greater than the size of the ob-
servable horizon. Inside our part of the universe, the sum
of all such perturbations can be interpreted as a homoge-
neous axion eld. This eld determines the initial value
of the axion eld at the onset of the axion oscillations,
and, as a result, it determines the ratio of dark matter
to usual matter in our universe [29]. If we follow only
the degrees of freedom inside our horizon, we may miss
this fact, as well as the possibility to explain the present
ratio of the dark matter to normal matter by anthropic
considerations [2931]. The same is true with respect to
some other eects which we mentioned in section II, such
as the possibility to give an anthropic explanation of the
baryon asymmetry of the observable part of the universe
in the Aeck-Dine scenario [27, 28] and the possibility to
explain the amplitude of perturbations of metric in the
curvaton scenario [32, 36].
One way to take into account this missing information
during the counting of all possible universes is to apply
the coarse-graining ideology. For example, during eter-
nal ination in the axion theory, the axion eld becomes
distributed all over the periodic phase space of its val-
ues, from 0 to 2f
a
, where f
a
is the radius of the axion
potential. In terms of the coarse-grained histories, this
dispersion may be represented as consisting of 4
2
f
a
/H
I
intervals of length
HI
2
. If this interval were in the range
of wavelengths within our horizon, it would contribute
an exponentially large factor to the number of possible
universes. Inside the horizon we do not have any informa-
tion about the exact history of perturbations with super-
horizon wavelength, but we still have a factor 4
2
f
a
/H
I
describing a family of coarse-grained possibilities for the
locally observable properties of the universe lled by a
classical oscillating axion eld. This extra factor is not
exponentially large, but if one ignores it, one could miss
one of the most interesting anthropic predictions of the
theory of inationary multiverse.
VI. ENTROPY OF MATTER
So far we were mainly interested in the distinct clas-
sical geometries produced by inationary perturbations
of metric. These perturbations provide the set of classi-
cal initial conditions for the subsequent evolution of the
universe. In our study we concentrated on perturbations
with an amplitude smaller than O(1), which produce lo-
cally Friedmann parts of the universe. For completeness,
we will briey discuss here the entropy of the usual mat-
ter, and also the entropy which can be produced when
perturbations of metric become large, which leads to
black hole production and their subsequent evaporation.
Various issues related to the contents of this section have
been discussed in many classical papers on cosmology, as
well as in more recent works including Refs. [53, 55].
First of all, let us estimate the maximal amount of
entropy of normal matter which can be accessible to an
observer in a universe with = O(1) and a cosmological
constant . The most important constraint here comes
from the fact that for > 0, particles leave the observ-
able part of dS space within the time t
1/2
, whereas
for < 0 the universe typically collapses within the time
t ||
1/2
, so in both cases in order to estimate the total
entropy produced after reheating that one can observe,
it is sucient to limit ourselves to what one can observe
within the cosmological time t ||
1/2
.
The total entropy of the universe will be maximized if
at the time when the energy density becomes comparable
8
with the cosmological constant, all particles are ultrarel-
ativistic. Assuming, for simplicity, that the number of
types of massless particles is O(1), one nds T
4
||
and the total entropy within a sphere of radius ||
1/2
is S
matter
||
3/4
. If the energy density at t ||
1/2
is not dominated by ultrarelativistic particles, the total
entropy of observable matter will be only smaller than
||
3/4
, so one has a bound
S
matter
||
3/4
, (35)
The same result can be obtained using the results of Ref.
[47], see [54]. This bound is very similar to the up-
per bound on the entropy of inationary perturbations
(26). However, one can obtain a more precise and strin-
gent bound for the entropy of inationary perturbations,
which depends not only on but also on the Hubble con-
stant at the end of ination: S
pert
H
3
2
I
||
3/4
(30).
In our universe, the upper bound S
matter
||
3/4
would yield the total entropy of particles S
matter
10
90
.
However, in reality the energy density of photons is sev-
eral orders of magnitude smaller than the energy density
of baryons, which is about 5 times smaller that the en-
ergy density of dark matter. Therefore the total entropy
of particles populating the observable part of our uni-
verse is several orders of magnitude smaller than its upper
bound O(||
3/4
): S
matter
10
88
. Thus, the main rea-
son why the upper bound is not exactly saturated lies in
the fact that ultrarelativistic matter (photons, gravitons
etc.) contribute only a small fraction to the total energy
density of the universe, as compared to baryonic matter
and dark matter, at the moment when this density drops
down to ||. It is interesting that the ratio of the energy
density of photons to energy density of nonrelativistic
matter may have an anthropic origin [28, 2831]. Thus,
anthropic considerations may explain the reason why the
upper bound on the entropy of particles is almost satu-
rated in our universe.
As we already mentioned before, the total entropy of
inationary perturbations in the observable part of our
universe is expected to be further from saturating the
upper bound ||
3/4
, see Eq. (31).
One should note, that once the perturbations of met-
ric grow and black holes form and evaporate, the total
entropy inside the observable part of the universe may
considerably increase. This is what happens in our uni-
verse, where the entropy of black holes in the centers of
galaxies can be greater than 10
104
[55]. In particular, the
entropy of a single black hole at the center of our own
galaxy can be greater than the entropy of all particles in
the observable part of the universe!
In the long run, most of the neighboring galaxies will
move further and further away from our galaxy due to
cosmic acceleration. If our galaxy (together with An-
dromeda) will eventually collapse into a single gigantic
black hole, its entropy will approach
S
MilkyHole
10
100
. (36)
Of course, the entropy will be much smaller if some parts
of matter in our galaxy form many smaller black holes
which will evaporate earlier. Moreover, it would take a
very long time for the Milky Hole to form and an even
longer time for us to observe its entropy in the form of
Hawking radiation. It is interesting, nevertheless, that
the total entropy produced by all localized objects in the
observable part of our universe can be totally dominated
by the entropy produced by the black hole evaporation.
If instead of considering our part of the universe we will
consider all regimes that are possible in the landscape,
one may envisage the possibility that the total entropy
of a cosmological black hole may approach the dS entropy
O(
1
). This may happen, for example, if the amplitude
of density perturbations on length scale ||
1/2
can be
O(1); see a related discussion in [56, 57].
VII. COUNTING WORLDS AND MINDS
In our calculations of the total number of dierent uni-
verses in the previous sections, we were assuming that
because the large scale uctuations of the scalar eld can
be interpreted as classical elds, all of the dierent uni-
verses produced by eternal ination have some kind of
real, observer-independent existence. However, each time
the meaning of these words was somewhat dierent.
When we were talking about all possible universes pro-
duced during eternal ination, we counted everything
that could be measured by all kinds of observers which
may live everywhere in the multiverse. In other words,
we counted all possible classical or semiclassical cong-
urations, all possible histories, not only the ones associ-
ated with the observable events inside the cosmological
horizon.
When we started talking about the universes inside the
horizon, we paid attention to the fact that the total num-
ber of outcomes which can be registered by any particular
observer at any moment of time is smaller than the total
number of possibilities which could emerge in all parts
of the universe. For example, an observer living inside
a horizon-size patch of an exponentially expanding uni-
verse does not have access to other parts of the universe.
Therefore some authors argue that anything that hap-
pens outside the horizon should not play any role in our
counting of the universes and evaluation of probabilities.
We do not want to discuss here validity of this argu-
ment. Instead of that, we would like to note that there
are additional quantum mechanical limitations on what
can be actually observed by any local observer. For ex-
ample, when one considers the Schrodinger cat exper-
iment, this experiment has two denite outcomes: the
cat can be either dead or alive. However, in accordance
with the Copenhagen interpretation, these potentialities
become realized only after one of these outcomes becomes
registered by a classical observer. In the many-world (rel-
9
ative state) interpretation of quantum mechanics, we are
talking about correlations between various observations
made by an observer and the state of the rest of the uni-
verse.
In everyday life, observers are big and very much classi-
cal, so their quantum nature can be safely ignored. How-
ever, the crucial ingredient of our procedure of counting
the universes was an investigation of quantum eects on
a wide range of scales from Planck length to supergalactic
scales. Meanwhile each of us is 10
26
times smaller than
the cosmological horizon and 10
35
times larger than the
Planck scale. Thus one may wonder to which extent one
can talk about a classical observer when discussing quan-
tum eects on the scales much smaller or much greater
than the size of an observer. Are there any constraints
on the total number of distinguishable universes which
are related to the quantum nature of an observer?
This issue becomes manifest when one remembers that
the essence of the Wheeler-DeWitt equation, which is
the Schrdinger equation for the wave function of the
universe, is that this wave function does not depend on
time, since the total Hamiltonian of the universe, includ-
ing the Hamiltonian of the gravitational eld, vanishes
identically [49].
The resolution of this paradox suggested by Bryce De-
Witt [49] is rather instructive. The notion of evolution is
not applicable to the universe as a whole since there is no
external observer with respect to the universe, and there
is no external clock that does not belong to the universe.
However, we do not actually ask why the universe as a
whole is evolving. We are just trying to understand our
own experimental data. Thus, a more precisely formu-
lated question is why do we see the universe evolving in
time in a given way. In order to answer this question one
should rst divide the universe into two main pieces. The
rst part consists of an observer with his clock and other
measuring devices, with a combined mass M and a total
energy Mc
2
. The second part is the rest of the universe,
with the total energy Mc
2
. Since the Hamiltonian (the
energy) of the rest of the universe does not vanish, the
wave function of the rest of the universe does depend on
the state of the clock of the observer, i.e. on his time.
One of the implications of this result is that one can
talk about the evolution of the universe only with respect
to an observer. In the limit when the mass of the observer
vanishes, the rest of the universe freezes in time. In this
sense, the number of distinct observable histories of the
universe is bounded from above by the total number of
the histories that can be recorded by a given observer.
And this number is nite.
Indeed, the total number N of all observable universes
which could be recorded by a given observer is bounded
from above by e
I
, where I is the maximal information
that he/she can collect. For any observer of mass M and
size R, this information cannot exceed the Bekenstein
bound
I < S
Bek
= 2MR. (37)
This bound implies that
N
observer
< e
S
Bek
= e
2MR
. (38)
For a typical observer with M 10
2
kg and R 1 m,
one nds
N
observer
e
10
45
. (39)
Moreover, if we consider a typical human observer, the
total amount of information he can possibly absorb dur-
ing his lifetime is expected to be of the order of 10
16
bits
or so [50]. In other words, a typical human brain can
have about
N
observer
10
10
16
(40)
dierent congurations, which means that a human ob-
server may distinguish no more than 10
10
16
dierent uni-
verses. This is a huge number, which is much greater
than the standard estimate of the number of dS vacua
in the landscape 10
500
. However, this number is much
smaller than the total number of possible geometries of
the universe inside the cosmological horizon after 60 e-
folds of ination.
Thus we are discussing an additional constraint which
previously did not attract much attention: The total
number of possibilities accessible to any given observer is
limited not only by the entropy of perturbations of metric
produced by ination and by the size of the cosmological
horizon, but also by the number of degrees of freedom of
an observer. This number is tremendously large, so one
can safely ignore this limitation in his/her everyday life.
But when we study quantum cosmology, evaluate the to-
tal number of the universes and eventually apply these
results to anthropic considerations, one may need to take
this limitation into account. Potentially, it may become
very important that when we analyze the probability of
existence of a universe of a given type, we should be talk-
ing about a consistent pair: the universe and an observer
who makes the rest of the universe alive and the wave
function of the rest of the universe time-dependent.
VIII. CONCLUSION
In this paper we made an attempt to nd out how
many dierent coarse-grained universes could be pro-
duced by ination in each particular vacuum, and in the
string theory landscape as a whole. The meaning of these
words can be explained as follows. Slow-roll ination pro-
duces long-wavelength perturbations of the metric, which
become imprinted on the cosmological background and
determine the large scale structure of the universe. Even
10
though these perturbations are created from quantum
uctuations, they become essentially classical due to in-
ation. These perturbations provide dierent classical
initial conditions in dierent parts of the universe. Our
goal was to estimate the number of distinctly dierent
classical geometries which may appear as a result of this
eect. We found that the result is proportional to e
e
3N
,
where N is the number of e-foldings of slow-roll ina-
tion. This aspect allows one to look from a dierent
perspective on the possible signicance of slow-roll ina-
tion, which helps to create the information content of the
universe.
The estimate of the total number of distinct geome-
tries produced by ination depends on the method by
which one can make this distinction. In the rst part of
this paper we concentrated on investigation of all possi-
ble locally Friedmann geometries which can be produced
after the end of eternal ination. Our goal was to under-
stand how many dierent locally-Friedmann (i.e. approx-
imately homogeneous and isotropic) universes constitute
the multiverse, which, as a whole, looks like a very in-
homogeneous and anisotropic non-Friedmann eternally
growing fractal. We found that the total number of such
universes, in the simplest inationary models, may ex-
ceed 10
10
10
7
. This humongous number is strongly model-
dependent and may change when one uses dierent de-
nitions of what is the boundary of eternal ination.
Then we decided to limit ourselves to only those uni-
verses which can be distinguished from each other by a
local observer in a universe with a given cosmological
constant . The resulting number appears to be limited
by e

3/4
. If this limit can be saturated, then the total
number of locally distinguishable congurations in string
theory landscape can be estimated by e
M
3/4
, where M
is the total number of vacua in string theory. In other
words, the total number of locally distinguishable geome-
tries is expected to be exponentially greater than the to-
tal number of vacua in the landscape.
Finally, we checked how many of these geometries can
be actually distinguished from each other by a local ob-
server of given mass and size. Not surprisingly, since any
local observer is smaller than the observable part of the
universe, we have found that the strongest limit on the
number of dierent locally distinguishable geometries is
determined mostly by our own abilities to distinguish be-
tween dierent universes and to remember our results.
In this paper we did not attempt to draw deep philo-
sophical conclusions based on our estimates, or apply
them immediately to the search for the probability mea-
sure in the multiverse. Just as those who calculated the
number of all possible vacua in the landscape, we con-
centrated on nding some facts, leaving their interpreta-
tion for further investigation. For example, it might be
worthwhile to explore some simple measures which could
emerge from our discussion. What would be the observa-
tional predictions if each of the universes have the prob-
ability to be observed P =
1
N
? What if the probability
is proportional to the observable entropy of inationary
perturbations P S? Is it possible to apply our results
to the stationary measure [23, 58]? We are planing to
return to these and other related issues in the future.
Acknowledgments
The authors are grateful to Jaume Garriga, Lev Kof-
man, Slava Mukhanov, Alex Vilenkin, Alexander West-
phal and Sergei Winitzki for helpful discussions. The
work of A. L. was supported in part by NSF grant PHY-
0244728, by the Alexander-von-Humboldt Foundation,
and by the FQXi grant RFP2-08-19. The work of V. V.
was supported in part by FQXi mini-grants MGB-07-018
and MGA-09-017.
[1] A.D. Linde, Nonsingular Regenerating Inationary Uni-
verse, Print-82-0554, Cambridge University preprint,
1982, see http://www.stanford.edu/alinde/1982.pdf
[2] A.D. Linde, The New Inationary Universe Sce-
nario, In: The Very Early Universe, ed. G.W.
Gibbons, S.W. Hawking and S.Siklos, Cam-
bridge University Press (1983), pp. 205-249, see
http://www.stanford.edu/alinde/1983.pdf
[3] A. Vilenkin, The Birth Of Inationary Universes, Phys.
Rev. D 27, 2848 (1983).
[4] A. D. Linde, Eternally Existing Self-reproducing
Chaotic Inationary Universe, Phys. Lett. B 175, 395
(1986).
[5] A. D. Linde, Particle Physics and Inationary Cosmol-
ogy, Physics Today 40, 61 (1987).
[6] S. Kachru, R. Kallosh, A. Linde and S. P. Trivedi, De
Sitter vacua in string theory, Phys. Rev. D 68, 046005
(2003) [arXiv:hep-th/0301240].
[7] M. R. Douglas, The statistics of string / M theory
vacua, JHEP 0305 046 (2003) [arXiv:hep-th/0303194];
F. Denef and M. R. Douglas, Distributions of ux
vacua, JHEP 0405, 072 (2004) [arXiv:hep-th/0404116];
M. R. Douglas and S. Kachru, Flux compactication,
[arXiv:hep-th/0610102]; F. Denef, M. R. Douglas and
S. Kachru, Physics of string ux compactications,
[arXiv:hep-th/0701050].
[8] W. Lerche, D. Lust and A. N. Schellekens, Chiral Four-
Dimensional Heterotic Strings from Selfdual Lattices,
Nucl. Phys. B 287, 477 (1987).
[9] R. Bousso and J. Polchinski, Quantization of four-
form uxes and dynamical neutralization of the
cosmological constant, JHEP 0006, 006 (2000)
[arXiv:hep-th/0004134].
[10] L. Susskind, The anthropic landscape of string theory,
11
arXiv:hep-th/0302219.
[11] J. Garriga and A. Vilenkin, Many worlds in one, Phys.
Rev. D 64, 043511 (2001) [arXiv:gr-qc/0102010].
[12] A.D. Linde and A. Mezhlumian, Stationary universe,"
Phys. Lett. B 307, 25 (1993) [arXiv:gr-qc/9304015].
[13] A.D. Linde, D.A. Linde and A. Mezhlumian, From the
Big Bang theory to the theory of a stationary universe,"
Phys. Rev. D 49, 1783 (1994) [arXiv:gr-qc/9306035].
[14] J. Garcia-Bellido, A.D. Linde and D.A. Linde, Fluctu-
ations of the gravitational constant in the inationary
Brans-Dicke cosmology," Phys. Rev. D 50, 730 (1994)
[arXiv:astro-ph/9312039].
[15] V. Vanchurin, A. Vilenkin and S. Winitzki, Predictabil-
ity crisis in inationary cosmology and its resolution,"
Phys. Rev. D 61, 083507 (2000) [arXiv:gr-qc/9905097.
[16] J. Garriga, D. Schwartz-Perlov, A. Vilenkin and S.
Winitzki, Probabilities in the inationary multiverse,"
JCAP 0601, 017 (2006) [arXiv:hep-th/0509184].
[17] V. Vanchurin and A. Vilenkin, Eternal observers and
bubble abundances in the landscape," Phys. Rev. D 74,
043520 (2006) [arXiv:gr-qc/0605015].
[18] V. Vanchurin, Geodesic measures of the landscape,"
Phys. Rev. D 75, 023524 (2007) [arXiv:hep-th/0612215].
[19] R. Bousso, Holographic probabilities in eter-
nal ination," Phys.Rev.Lett. 97, 191302 (2006)
[arXiv:gr-qc/0605263].
[20] A. Linde, Towards a gauge invariant volume-weighted
probability measure for eternal ination," JCAP 0706,
017 (2007) [arXiv:0705.1160[hep-th]].
[21] S. Winitzki, A Volume-weighted measure for eter-
nal ination," Phys. Rev. D 78, 043501 (2008)
[arXiv:0803.1300[gr-qc]].
[22] J. Garriga and A. Vilenkin, Holographic Multiverse,"
JCAP 0901, 021 (2009) [arXiv:0809.4257[gr-qc]].
[23] A. Linde, V. Vanchurin and S. Winitzki, Stationary
Measure in the Multiverse," JCAP 0901, 031(2009)
[arXiv:0812.0005[hep-th]]
[24] D.N. Page, The Born Rule Dies," JCAP 0907, 008
(2009) [arXiv:0903.4888[hep-th]]
[25] V. F. Mukhanov and G. V. Chibisov, Quantum Fluc-
tuation And Nonsingular Universe, JETP Lett. 33,
532 (1981) [Pisma Zh. Eksp. Teor. Fiz. 33, 549 (1981)];
S. W. Hawking, The Development Of Irregularities In A
Single Bubble Inationary Universe, Phys. Lett. B 115,
295 (1982); A. A. Starobinsky, Dynamics Of Phase Tran-
sition In The New Inationary Universe Scenario And
Generation Of Perturbations, Phys. Lett. B 117, 175
(1982); A. H. Guth and S. Y. Pi, Fluctuations In The
New Inationary Universe, Phys. Rev. Lett. 49, 1110
(1982); J. M. Bardeen, P. J. Steinhardt and M. S. Turner,
Spontaneous Creation Of Almost Scale - Free Density
Perturbations In An Inationary Universe, Phys. Rev.
D 28, 679 (1983); V. F. Mukhanov, Gravitational In-
stability Of The Universe Filled With A Scalar Field,
JETP Lett. 41, 493 (1985) [Pisma Zh. Eksp. Teor. Fiz.
41, 402 (1985)]; V. F. Mukhanov, Physical Foundations
of Cosmology, Cambridge University Press, 2005.
[26] L. Kofman, A. D. Linde and A. A. Starobinsky, Re-
heating after ination, Phys. Rev. Lett. 73, 3195
(1994) [arXiv:hep-th/9405187]; L. Kofman, A. D. Linde
and A. A. Starobinsky, Towards the theory of re-
heating after ination, Phys. Rev. D 56, 3258 (1997)
[arXiv:hep-ph/9704452].
[27] I. Aeck and M. Dine, A New Mechanism For Baryoge-
nesis, Nucl. Phys. B 249, 361 (1985).
[28] A. D. Linde, The New Mechanism Of Baryogenesis
And The Inationary Universe, Phys. Lett. B 160, 243
(1985).
[29] A. D. Linde, Ination And Axion Cosmology, Phys.
Lett. B 201, 437 (1988).
[30] M. Tegmark, A. Aguirre, M. Rees and F. Wilczek,
Dimensionless constants, cosmology and other
dark matters, Phys. Rev. D 73, 023505 (2006)
[arXiv:astro-ph/0511774].
[31] B. Freivogel, Anthropic Explanation of the Dark Matter
Abundance, arXiv:0810.0703 [hep-th].
[32] A. D. Linde and V. Mukhanov, Nongaussian isocurva-
ture perturbations from ination, Phys. Rev. D 56, 535
(1997) [arXiv:astro-ph/9610219].
[33] K. Enqvist and M. S. Sloth, Adiabatic CMB perturba-
tions in pre big bang string cosmology, Nucl. Phys. B
626, 395 (2002) [arXiv:hep-ph/0109214].
[34] D. H. Lyth and D. Wands, Generating the curvature
perturbation without an inaton, Phys. Lett. B 524, 5
(2002) [arXiv:hep-ph/0110002].
[35] T. Moroi and T. Takahashi, Eects of cosmological mod-
uli elds on cosmic microwave background, Phys. Lett.
B 522, 215 (2001) [Erratum-ibid. B 539, 303 (2002)]
[arXiv:hep-ph/0110096].
[36] A. D. Linde and V. Mukhanov, The curvaton web,
JCAP 0604, 009 (2006) [arXiv:astro-ph/0511736].
[37] R. H. Brandenberger, V. F. Mukhanov and T. Prokopec,
Entropy of a classical stochastic eld and cosmolog-
ical perturbations," Phys. Rev. Lett. 69, 3606 (1992)
[arXiv:astro-ph/9206005].
[38] R. H. Brandenberger, T. Prokopec and V. F. Mukhanov,
Entropy of the Gravitational Field," Phys. Rev. D 48,
2443 (1993) [arXiv:astro-ph/9208009].
[39] C. Kiefer, D. Polarski and A. A. Starobinsky, Entropy
of gravitons produced in the early universe, Phys. Rev.
D 62, 043518 (2000) [arXiv:gr-qc/9910065].
[40] D. I. Podolsky, G. N. Felder, L. Kofman and M. Peloso,
Equation of state and beginning of thermalization
after preheating, Phys. Rev. D 73, 023501 (2006)
[arXiv:hep-ph/0507096].
[41] C. Kiefer and D. Polarski, Why do cosmological pertur-
bations look classical to us?, arXiv:0810.0087 [astro-ph].
[42] D. Campo and R. Parentani, Decoherence and entropy
of primordial uctuations II. The entropy budget, Phys.
Rev. D 78, 065045 (2008) [arXiv:0805.0424 [hep-th]].
[43] A. Linde, Particle Physics and Inationary Cos-
mology," Harwood, Chur, Switzerland (1990)
[arXiv:hep-th/0503203].
[44] A. Linde and A. Westphal, Accidental Ination in String
Theory, JCAP 0803, 005 (2008) [arXiv:0712.1610 [hep-
th]].
[45] S. Winitzki, Observability of the total inationary ex-
pansion, arXiv:1003.1680 [gr-qc].
[46] R. Bousso, B. Freivogel,S. Leichenauer, "Saturating the
holographic entropy bound", [arXiv:1003.3012 [hep-th]].
[47] G. t Hooft, Dimensional reduction in quantum gravity,
Salamfest 0284-296 (1993) arXiv:gr-qc/9310026.
[48] N. Arkani-Hamed, S. Dubovsky, A. Nicolis,
E. Trincherini and G. Villadoro, A Measure of de
Sitter Entropy and Eternal Ination, JHEP 0705, 055
(2007) [arXiv:0704.1814 [hep-th]].
[49] B. S. DeWitt, Quantum Theory of Gravity. 1. The
Canonical Theory, Phys. Rev. 160, 1113 (1967).
12
[50] A. De Simone, A. H. Guth, A. Linde, M. Noor-
bala, M. P. Salem and A. Vilenkin, Boltzmann brains
and the scale-factor cuto measure of the multiverse,
[arXiv:0808.3778 [hep-th]].
[51] L. Susskind, The Census Takers Hat,"
[arXiv:0710.1129[hep-th]]
[52] R. Kallosh and A. D. Linde, Landscape, the scale of
SUSY breaking, and ination, JHEP 0412, 004 (2004)
[arXiv:hep-th/0411011].
[53] R. Bousso, R. Harnik, G. D. Kribs and G. Perez,
Predicting the Cosmological Constant from the Causal
Entropic Principle, Phys. Rev. D 76, 043513 (2007)
[arXiv:hep-th/0702115].
[54] R. Bousso, a talk at KITP Conference String Phe-
nomenology 2006, unpublished.
[55] P. H. Frampton and T. W. Kephart, Upper and
Lower Bounds on Gravitational Entropy, JCAP 0806,
008 (2008) [arXiv:0711.0193 [gr-qc]]; C.A.Egan, C.H.
Lineweaver, A Larger Estimate of the Entropy of the
Universe," [arXiv:0909.3983 [astro-ph.CO]]; P. H. Framp-
ton and K. Ludwick, Number and Entropy of Halo Black
Holes, arXiv:0910.1152 [astro-ph.GA].
[56] R. Bousso and S. W. Hawking, Pair creation of black
holes during ination, Phys. Rev. D 54, 6312 (1996)
[arXiv:gr-qc/9606052].
[57] J. Garcia-Bellido, A. D. Linde and D. Wands, Den-
sity perturbations and black hole formation in hy-
brid ination, Phys. Rev. D 54, 6040 (1996)
[arXiv:astro-ph/9605094].
[58] A. D. Linde, Towards a gauge invariant volume-weighted
probability measure for eternal ination, JCAP 0706,
017 (2007) [arXiv:0705.1160 [hep-th]].